Headphone Sensitivity & Power Requirements

[Ken Tajalli]

I have had some heated discussions (disagreements!) with a fellow member regarding this issue. I believe it deserves a discussion of its own.
If there are similar threads, or it has already been discussed to death, I apologize, please point me to them.

There are sensitivity values given for headphones, it is given as dB for 1mW or dB for 1V RMS.
The general attitude, and a mathematical one! Is, to use this figure and calculate, how much power is needed to achieve a healthy loudness of say 115dB. Then we can safely say, this headphone needs an amp that can produce this much power.
Straight forward, makes sense! No?

I say No!
And that is the gist of my disagreement with my fellow member.
My arguments against:
- First see this

I have stated among my many arguments previously, that measurements for Electronic devices (DACs, Amps) can be absolute, but when it comes to Transducers, it is not.
Firstly we should recognize that, the sensitivity values given, are just an indication-tool, used for comparison.
Take two headphones for example, one has a dip in its FR curve at 400Hz, the other has a peak at the same frequency. Do you think if I pick 400Hz to measure the sensitivity, the value I get, is going to be reliable?
Secondly, in the scope of calculating how much power is needed, Loudness is a subjective value.
It depends on the type of music being played, the FR of the DUT, aaaand . . . , the hearing of the listener's ears.
Power and Gain are also inter-related. Higher power amps have more gain, and sometimes, we may need the gain, and not necessarily the max. power. With headphones, people equalize. If you equalize Bass frequencies, you need to use pre-cuts, right?
The amp being used, can quickly run out of power at bass, and overall loudness suffers. Now what if we didn't EQ the bass, but we EQ the mids up? Now we don't need to use a pre-cut, Our ears are sensitive to the mids, so the headphones will become louder with the same power capability.
Finally, someone with a failing hearing, will need more power to Hear better.
So I state that calculating (guessing) power requirement for headphones, is not as straight forward as some believe. It is complex, with variables depending on case to case surroundings.
What do you think?

 

[DVDdoug]

IMO, the bigger issue is that power also depends on impedance. Headphone impedance has a very wide range and a headphone amp that can put-out 1W at 32 Ohms may not be able to put-out 1W at 300 Ohms. With a power amplifier it's easier because you can specify power at 4 & 8 Ohms.

I'd prefer a voltage spec.

I believe sensitivity is usually specified at 1kHz.

 

[solderdude]

Take two headphones for example, one has a dip in its FR curve at 400Hz, the other has a peak at the same frequency. Do you think if I pick 400Hz to measure the sensitivity, the value I get, is going to be reliable?

No it will not be useful but also not useless as the difference will likely just be a few dB max.
Add to that the fixture must be calibrated as well. Add to that product tolerance/variance.

Also know that manufacturers sensitivity and or efficiency numbers often are incorrect. Yes, also from the big manufacturers.
Furthermore it is mostly unknown at which frequency (or frequency band) it is measured and on what gear and if this was a prototype or production type.
Often just dB is mentioned or dBmW is listed but somehow a dBV value is used (or v.v.).
It is a mess.

To calculate needed 'power' this number needs to be correct.

Secondly, in the scope of calculating how much power is needed, Loudness is a subjective value.
It depends on the type of music being played, the FR of the DUT, aaaand . . . , the hearing of the listener's ears.

Yes, depends on DR, and how loud it is recorded and the frequency response of the headphone as well as the applied EQ (if needed).
To calculate the power you will also need to know the impedance.
Better to use dBV if you want to compare headphones and calculations.

I built myself a handy meter that, when one knows the sensitivity (dBV) can instantly calculate peak SPL and see what peak voltages were present.
The meter is very fast and has 'peak hold'.

I gauge the sensitivity of a headphone by looking at the 'average' level of mids (and a bit of the bass) and for reference I use a HD650 of which I know the sensitivity and know which voltage is applied.

Due to the 'wobbly' nature of a headphone's frequency response it may not be the best idea to measure at 400Hz or 1kHz only as there may be dips or elevations at that point relative to the 'average' SPL.

 

[Ken Tajalli]

No it will not be useful. Add to that the fixture must be calibrated as well. Add to that product tolerance/variance.


Yes, depends on DR, and how loud it is recorded and the frequency response of the headphone as well as the applied EQ (if needed).
To calculate the power you will also need to know the impedance.
Better to use dBV if you want to compare headphones and calculations.

I built myself a handy meter that, when one knows the sensitivity (dBV) can instantly calculate peak SPL and see what peak voltages were present.
The meter is very fast and has 'peak hold'.
View attachment 390226

Excellent!
Also not forgetting that not every amp can drive low impedance loads. Some may clip or go into protection given a low impedance.
So even a dBV should be viewed in that regard.

 

[NTK]

Excellent!
Also not forgetting that not every amp can drive low impedance loads. Some may clip or go into protection given a low impedance.
So even a dBV should be viewed in that regard.

Amir's charts of distortion+noise vs output voltage for a range of loads give that information.

 

[solderdude]

Specs should best be given in dBV where it is meant the measured SPL when 1V is applied.

Fortunately, after calculating the found voltage of Amir's measurements (400Hz, 94dB SPL) to dBV I often find his measurements are largely in agreement with mine (within tolerance range).

 

[Joachim Herbert]

On a side note, 115 dB is all but a healthy level.

 

[Ken Tajalli]

On a side note, 115 dB is all but a healthy level.

Sort of!
115dB peak is about 95dB or so, average listening.

 

[solderdude]

115dB momentary peaks in music with a DR>15 and only for a song is something entirely different than an average noise level of 115dBA.

The thing is that IF one every now and then likes to turn up a song really loud for a short moment and don't want any clipping it is best to have an amplifier that can supply enough voltage to reach that level.
It is also a good idea to test headphones in the lowest frequencies at that level to see how well they respond to that.

For instance... the FiiO FT1 is 117dBV and a (non E.U.) phone can supply 1V. Most dongles even can supply 2V.
Putting it on my head and listening to 'all evening relaxing levels' in other words soft level I measure peaks of 30mV (87dB peaks, with DR12 song = 75dB average).
Listening to more 'active' listening levels (during the day trying to evaluate sound quality) I measure 200mV peaks (102dB peaks)
Listening to loud levels that are 'tolerable' for a minute or so I measure 700mV peaks (114dB peaks with DR17 recording = 97dB average, dBA will be a lower value).

My HD800 (with EQ) requires about 14dB (5x) higher output voltage for the same listening levels.

 

[oleg87]

Simply factor in a healthy margin to your listening volume so you don't have to worry about the vagaries of sensitivity measurements.

 

[Ken Tajalli]

Simply factor in a healthy margin to your listening volume so you don't have to worry about the vagaries of sensitivity measurements.

"Margin" to what?
That's the question!
My LCD-XC connected to my Hugo2 has a healthy margin, with a Susvara?, that would be a different story.
Headphone power, specially on portable devices, comes at a premium.